I used the
spls R-package to select relevant variables from a list of 1600 continuous predictor variables (genes). The response variable is expression of another gene (continuous).
After running the
spls function as follows I get a variable importance score for each predictor variable. I can sort from highest to lowest and I see some of the important ones have higher score (so the method works). But how do I do a cutoff in a such a way that is statistically significant?
any suggestion would be really helpful.
library(spls) cv <-cv.spls(Xvars,Yvar.lignin.exp,eta = seq(0.1,0.9,0.1),K = c(5:10), kappa=0.5, select="pls2", fit="simpls",scale.x=TRUE, scale.y=TRUE, plot.it=F) f <- spls(Xvars,Yvar,eta = cv$eta.opt,K = cv$K.opt) coef.f <- coef(f) boolf<-coef.f > 0 boolf<-boolf*1 sparse.coeff<-coef.f*boolf sparse.coeff.stem <- data.Normalization (sparse.coeff,type="n4",normalization="column")